Deviance

Calculates the deviance.

Usage

.deviance(y, y_hat, family)

Arguments

y

response vector of length \(n_0\), containing numerical values (family="gaussian"), integer values (family="poisson"), binary values (family="binomial"), or a survival object created with survival::Surv() (family="cox"), where \(n_0\) is the number of observations used for model training

y_hat

predicted response: numeric vector of length \(n\), with entries on the real range (family="gaussian" or family="cox"), in the unit interval (family="binomial"), or on the non-negative real range (family="poisson")

family

character string "gaussian", "binomial", "poisson", or "cox"

Value

Returns the deviance (a numeric scalar).

See also

The function deviance.cv.corila() extracts the deviance from a fitted model.

Examples

n <- 10L

y <- rnorm(n)
y_hat <- rnorm(n)
.deviance(y = y, y_hat = y_hat, family = "gaussian")
#> [1] 1.73779

y <- rbinom(n = n, size = 1L, prob = 0.5)
y_hat <- runif(n)
.deviance(y = y, y_hat = y_hat, family = "binomial")
#> [1] 2.005729

y <- rpois(n = n, lambda = 4.0)
y_hat <- rexp(n)
.deviance(y = y, y_hat = y_hat, family = "poisson")
#> [1] 8.054846